Charles Spinelli on Making AI Explanations Useful at Work
Transparency is often treated as the answer to concerns about artificial intelligence in the workplace. When employees question automated recommendations, organizations may respond by providing more detail on data sources, model logic, scoring methods, and decision pathways. The intention is reasonable. People are more likely to trust systems they can understand. Charles Spinelli points out that transparency loses value when explanations become too dense, technical, or difficult to apply.
In practice, more information does not always lead to greater clarity. Employees may receive lengthy disclosures, complicated dashboards, or technical summaries that explain a system but do not help them interpret a decision. When explanation becomes a burden, transparency can shift from a tool for trust to another source of confusion.
When More Detail Becomes Less Helpful
AI systems can be difficult to explain because their outputs often come from many interacting inputs. A recommendation may reflect historical data, behavioral patterns, weighted variables, and system rules. Providing every detail may seem thorough, but it can overwhelm the people who are expected to use the information.
The issue is not whether organizations should explain AI decisions. The issue is whether those explanations are usable. A manager reviewing a performance alert or an employee questioning a scheduling recommendation needs practical context, not a technical report that requires specialized training. When explanations are overly complex, employees may ignore them or resort to simplified assumptions.
The Trust Problem Hidden Inside Complexity
Organizations may assume that detailed explanations strengthen trust. That assumption can fail when employees cannot connect the explanation to the outcome that affects them. A long description of model behavior may not answer a basic workplace question: Why did this decision happen to me?
Trust is not built solely on the volume of information. It depends on whether employees can see how the information relates to their work, their performance, or their opportunities. When transparency feels procedural rather than meaningful, employees may view it as a compliance exercise. The organization may technically provide information, but the employee may still feel left without a clear answer.
Designing Explanations Around Human Use
Useful transparency starts with the person receiving the explanation. Employees, managers, HR teams, and technical reviewers may need different levels of detail. A single explanation format rarely serves all audiences well.
Plain-language summaries can help employees understand the general basis of a decision. Managers may need more detail about confidence levels, data limits, and points where human review is appropriate. Technical teams may require deeper access to model behavior and system performance. Layered explanations can help reduce fatigue.
Turning Transparency into Understanding
AI explanations should help people act, question, or respond. That requires clarity about what the system considered, what it did not consider, and where human judgment remains relevant. Without those elements, transparency may satisfy a formal requirement while leaving employees in the dark. Organizations can strengthen understanding by testing explanations with the people who use them.
Transparency has value only when it supports interpretation. Charles Spinelli points out that this is an important distinction for workplaces using AI in decisions that affect people directly. As automated systems become more common, organizations need explanations that inform rather than overwhelm. Clear, usable communication helps employees understand AI-supported decisions without turning transparency itself into another workplace burden.